DeepSeek V4.1 Flash 552B · Chip comparison

DeepSeek V4.1 Flash 552B — MI355X vs Vera Rubin NVL72

Head-to-head AI inference benchmark comparison of MI355X (AMD CDNA 4) and Vera Rubin NVL72 (NVIDIA Vera Rubin) on DeepSeek V4.1 Flash 552B. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.

AgentX replays real coding-agent sessions rather than fixed-length prompts, so context grows turn over turn and most of each request is served from cache instead of being recomputed. That turns the comparison into a systems question: KV transfer between nodes, prefix-aware routing, and cache capacity all move the curve alongside raw chip throughput. Fixed-sequence workloads stay the clean baseline for kernel and silicon performance, so the two scenarios answer different questions about the same hardware. Learn more about AgentX →

Near the low end of the 181–1333 tok/s/user interactivity band, at 469 tok/s/user on DeepSeek V4.1 Flash 552B: MI355X runs 30,320 tok/s/chip at $0.014 per million tokens, Vera Rubin NVL72 runs 1,439,777 at $0.001. Vera Rubin NVL72 is 95% cheaper per token; Vera Rubin NVL72 delivers 47× the tok/s/chip.

Setting 757 tok/s/user as the target on DeepSeek V4.1 Flash 552B, MI355X produces 3,829 tok/s/chip ($0.11 per million tokens) and Vera Rubin NVL72 produces 325,330 ($0.003). Vera Rubin NVL72 is 97% cheaper per token; Vera Rubin NVL72 delivers 85× the tok/s/chip.

At 1045 tok/s/user interactivity on DeepSeek V4.1 Flash 552B, MI355X delivers 3,158 tok/s/chip at $0.13 per million tokens; Vera Rubin NVL72 delivers 108,999 tok/s/chip at $0.009. Vera Rubin NVL72 is 93% cheaper per token; Vera Rubin NVL72 delivers 35× the tok/s/chip at this point. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)

View performance per dollar →

Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.

Metric
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Throughput (tok/s/chip)
MI355X30,320Vera Rubin NVL721,439,777
MI355X3,829Vera Rubin NVL72325,330
MI355X3,158Vera Rubin NVL72108,999
Cost ($/M tok)
MI355X$0.014Vera Rubin NVL72$0.001
MI355X$0.11Vera Rubin NVL72$0.003
MI355X$0.13Vera Rubin NVL72$0.009
tok/s/MW
MI355X14,507,115Vera Rubin NVL72436,295,912
MI355X1,832,092Vera Rubin NVL7298,584,793
MI355X1,510,963Vera Rubin NVL7233,030,100
Concurrency
MI355X~7Vera Rubin NVL72~9,783
MI355X~1Vera Rubin NVL72~1,213
MI355X~1Vera Rubin NVL72~256